MOBIL-based traffic prediction and interaction-aware Model Predictive Control for autonomous highway driving

This paper proposes an interaction-aware Model Predictive Control (MPC) approach for autonomous highway driving, introducing a novel framework to model vehicle interactions. First, the possible lateral motion behaviors of the autonomous vehicle and surrounding vehicles are predicted using the rule-based Minimizing Overall Braking Induced by Lane Change (MOBIL) model, which evaluates actions based on their overall benefit to traffic flow. These predicted behaviors are then categorized according to the lateral motion of the autonomous vehicle. Based on this categorization, different MPC control modes are developed for each category. Finally, by solving the MPC optimization problems for all control modes and selecting the one that minimizes the overall cost for all vehicles, the lateral motion decision of the autonomous vehicle is determined. In each control mode, the autonomous and surrounding vehicles interact longitudinally to ensure collision avoidance, while simultaneously considering traffic rules and driving comfort. The proposed controller is validated in a high-fidelity IPG CarMaker and Simulink co-simulation environment across diverse cases, as well as in a Monte Carlo simulation study. Results show that the autonomous vehicle can perform safely and improve traffic flow by changing lanes when necessary. Monte Carlo simulations further demonstrate the robustness and generality of the proposed method across various traffic conditions. • Use the MOBIL model to capture the lateral motion uncertainties of the EV and TVs. • Combine and optimize the longitudinal states of the EV and TVs using interaction-aware MPC. • Leverage a compact control structure to ensure efficiency via linear optimization. • Validate the method in diverse cases with the high-fidelity IPG CarMaker platform.